PROJECT SUMMARY The overall goal of this K01 application is to optimize clinical care decisions for people living with HIV. Specifically, this project will explore how cause-specific mortality among people with HIV has changed as treatment has become more effective and how the choice of antiretroviral therapy (ART) regimen can be tailored or personalized based on patient characteristics to improve survival. Since 2012, many patients have initiated regimens containing integrase inhibitors, but overall and cause-specific mortality for patients on these regimens is uncertain. In addition, the comparative effectiveness of the recommended integrase inhibitor containing regimens for patients with disparate characteristics and treatment histories has yet to be explored. Standard epidemiologic methods are insufficient to optimize HIV treatment plans because treatment plans and tailoring strategies are high dimensional, resulting in sparse data and unstable inference in many data sources, particularly when treatment plans can change over time. The goal of this career development project is to train the recipient to perform comparative effective research in settings with many exposure plans and outcomes. Research aims of this project are to 1) Compare the cause-specific mortality risks among patients with HIV in the US across three time periods representing the triple drug therapy era, the single tablet era, and the integrase inhibitor era (i.e., 2000 ? 2005, 2006 ? 2012, 2013 ? 2018); and 2) Estimate all-cause mortality risks under strategies to optimize selection of an integrase inhibitor containing regimen based on treatment history and patient characteristics. To address these aims in cohort data, the training component of this grant focuses on building expertise in semi-Bayesian semiparametric inference in the context of HIV research. Specifically, training aims include 1) Instruction in statistical techniques to improve inference for tailored treatment plans in high dimensional settings; 2) Training in applied HIV epidemiology; and 3) Experience and preliminary results necessary to prepare an R01 application in the fourth year of this award. The training aims will be achieved through rigorous coursework in advanced biostatistics, mentored and collaborative research, and conference participation. Research aims will be conducted using data from the Centers for AIDS Research Network of Integrated Clinical Systems, which includes over 30,000 HIV-seropositive adults engaged in clinical care from January 1, 1995 to the present at 8 US sites. The project will use semiparametric methods to account for missing causes of death and will estimate all parameters describing cause-specific mortality accounting for competing causes of death. Aim 2 will use Bayesian penalization techniques to estimate the causal effects of tailored treatment plans using marginal structural models and the parametric g-formula. This project will address an urgent need to optimize treatment plans in the current treatment era with an aging HIV-positive population with increasing comorbidities.